Join your knowledge with our AI models!
Portuguese DSO is using this tool daily for 1 year with 85% of Accuracy
Data Collection
Information such as: age, materials, lenght, and reliability.
Terrain characteristics like vegetation, protected areas, geographic relief and watercourse.
Number of lines in fault in a 3-hour period, on the outage management system.
Simulation of the last 5 years of all meteorological variables: wind speed and direction, gust, temperature, rainfall, lightning and atmospheric pressure.
Data Analysis
Transform the power lines into a discrete point system, from 100 to 100 meters.
This step has a great importance because it allows the synchronization of all the variables collected, in latitude and longitude.
This step has a great importance because it allows the synchronization of all the variables collected, in latitude and longitude.
Due to the small amount of data available, it was necessary to merge the knowledge of the technical team with the fault events, creating a global model that allows a line-to-line model, where the output is a risk index model of probability of fault.
Neural networks with all variables of historical data and meteorology.
Data Visualization
- Risk index for rain, wind, lightning and global.
- Number of incidents of short (<3min) and long (>3 min) term.
- Probabilistic distribution for long-term incidents.
- Number (power) of critical customers by zone.
- Accumulated precipitation, relative humidity and discharge indicator.
- Historical evaluation | expected incidents vs. actual incidents.
- Integration of weather maps | windy.
IS IT ACCURATE?
This tool is being used in Portugal for 1 year, with 80 000 km of power lines, and for 85% of the time the forecast matched with what happened in real time. It is also relevant to mention that for 99% of the time, when there was an outage, the real events happened in the forecasted level or above.
Take a look!
Get to know better our platform. It includes different widgets to fulfill all your needs.
Here's how it helped others
Outage Forecast is being used by EDP Distribuição, Portuguese DSO, for over one year now. It is an essential too for their grid daily planning.
'Outage Forecast was one of the first successful applications of Machine Learning techniques to an EDP Distribuição's real challenge. The deployment of this project help us better understand the impact of weather conditions across all mainland Portuguese HV and MV grid, and by predicting the number of outages help us improving our workforce management strategies and operacional planning'